DBAsupport.com Forums - Powered by vBulletin
Page 1 of 2 12 LastLast
Results 1 to 10 of 14

Thread: Configuration on XP 3 drives?

  1. #1
    Join Date
    Nov 2000
    Location
    Baltimore, MD USA
    Posts
    1,339

    Configuration on XP 3 drives?

    Okay, we all know how clueless I am with the administration stuff. Imagine a client that knows even less.

    Given this:
    -----------
    Windows XP
    4 processors
    4 Gig Memory
    3 135G Hard Drives
    1 instance running a small (<20G, <100 tables, <10 million total rows) OLTP database

    How should this thing be configured?

    I'm not looking for 'it depends', because the configuration they have is terrible and any 'average' setup would be better:

    How should the disks be striped/mirrored?
    Where should the os/data/redo/etc be located?
    What should the main memory params be?

    Again - I'm not looking for anything perfect here at all - just a general starting point for all this stuff.

    Thanks,

    - Chris
    Christopher R. Long
    ChrisRLong@HotMail.Com
    But that's just my opinion. I could be wrong

  2. #2
    Join Date
    Nov 2000
    Location
    greenwich.ct.us
    Posts
    9,092
    Three drives, ugh. Four would have made it easy; two mirrored pairs, one side of your redo logs on one pair, the other side on the second pair, spread your data out on both pairs evenly.

    Three is a little more difficult. I wouldn't want to setup a RAID 5 because the OS would have to be on it (swapping). If this system is not critical, I might put the OS on disk0, multiplex the redo logs on disk1 and disk2, spread the data on disk1 and disk2 and put your temp on disk0.

    If this is a critical system, I'd mirror a pair and put the OS, one copy of the redo logs, archived redo logs, undo, and data on the mirrored drive. On the other drive I would put indexes, other side of the redo logs, temp, and data that can be lost.

    Either case, I'd make sure my backup plan is SOLID.

    SGA, I'd start at about 512M in dedicated server mode and tune up from there (which you probably wouldn't need to do).
    Jeff Hunter

  3. #3
    Join Date
    Nov 2002
    Location
    Geneva Switzerland
    Posts
    3,142
    The client obviously does know MUCH less - how else would he buy 4 processors but only 3 disks! Can he swap them for 6 x 36GB? Or even 8 x 18GB? Much more useful.

    With so few disks you have a very difficult choice between fault tolerence and efficiency.

    My 2ยข about disks:
    - DON'T stripe - striping maximises the effect of disk failure.
    - If you want disk redundancy you could partition them as 2x65GB and software mirror them: 1a:2b 2a:3b 3a:1b (Windows Disk Manager does this - tho' its name might be different in XP!). This must have some overhead.
    - I think if you don't mirror, you MUST multiplex your logs over two disks. Make sure the paging file is on the third (probably with the OS). Spread everthing else around. Of course device separation is incompatible with mirroring as above.

    As for memory, I think your experience with *N*X would be as good a guide as anything.

  4. #4
    Join Date
    Nov 2000
    Location
    greenwich.ct.us
    Posts
    9,092
    Originally posted by DaPi
    The client obviously does know MUCH less - how else would he buy 4 processors but only 3 disks! Can he swap them for 6 x 36GB? Or even 8 x 18GB? Much more useful.

    I agree. more disks = better (probably)


    - If you want disk redundancy you could partition them as 2x65GB and software mirror them: 1a:2b 2a:3b 3a:1b (Windows Disk Manager does this - tho' its name might be different in XP!). This must have some overhead.
    The only thing you have to worry about here is that if one of the disks goes out, you now have both of the remaining drives without any redundancy.
    Jeff Hunter

  5. #5
    Join Date
    Nov 2002
    Location
    Geneva Switzerland
    Posts
    3,142
    Originally posted by marist89
    The only thing you have to worry about here is that if one of the disks goes out, you now have both of the remaining drives without any redundancy.
    At least it gives you time to worry The same is true for RAID-5. With RAID-1 plus no-RAID you have a 1-in-3 chance of panicking instead of just worrying when a disk goes AWOL.

    Another consideration is time to recover. Losing the OS is probably the slowest to recover from, since you need to install at least a mini-OS before you can start the restore process. Which supports putting the OS on the mirror in the RAID-1 plus no-RAID configuration.

    Hell, there is no good answer to this.

    P.S. I am told, by a SA type, that soft mirroring is more efficient than RAID-5. This could be lie . . . :(
    Last edited by DaPi; 09-24-2004 at 03:01 PM.

  6. #6
    Join Date
    Nov 2000
    Location
    greenwich.ct.us
    Posts
    9,092
    Originally posted by DaPi

    Hell, there is no good answer to this.

    agreed.


    P.S. I am told, by a SA type, that soft mirroring is more efficient than RAID-5. This could be lie . . . :(
    Sure a software RAID 1 (mirror) is more efficient than a software RAID 5. Software RAID 1 vs. Hardware RAID 5, depends on frequency of writes.
    Jeff Hunter

  7. #7
    Join Date
    Nov 2002
    Location
    Geneva Switzerland
    Posts
    3,142
    The SA was comparing soft RAID-1 with hard RAID-5.

    With so little thought gone into the disk config of this machine, I suppose there won't be a RAID controller so soft mirroring it is.

  8. #8
    Join Date
    Nov 2000
    Location
    Pittsburgh, PA
    Posts
    4,166
    I know this is silly, but I will say it anyway. Why don't they just buy more disks and a couple of good raid controllers?

    They could at least set up 3-4 raid 1 paritions based on a hardware level raid.

  9. #9
    Join Date
    Nov 2000
    Location
    greenwich.ct.us
    Posts
    9,092
    Originally posted by gandolf989
    I know this is silly, but I will say it anyway. Why don't they just buy more disks and a couple of good raid controllers?

    They could at least set up 3-4 raid 1 paritions based on a hardware level raid.
    I can hear it now..."But you said you wanted 400G of space."
    Jeff Hunter

  10. #10
    Join Date
    Nov 2000
    Location
    Pittsburgh, PA
    Posts
    4,166
    Originally posted by marist89
    I can hear it now..."But you said you wanted 400G of space."
    Of course as you know its about the response time and the reliability. Maybe having 400G of diskspace is what it takes to store less than 30G total worth of stuff. I recently worked as a DBA at a site with 5TB worth of disk drives for 1TB of data because they wanted a minimum number of 10,000 rpm disks and thats what they ended up with. They would have bought 18 gig drives but couldn't get enough of them so they ended up with 36 gig drives instead, they bought the same number of drives regardless. Disk space is relatively cheap.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  


Click Here to Expand Forum to Full Width